The 2-day cadence: why our own blog slipped 14 days
The irony of automation: tamer promises a breakneck delivery cadence, but we missed our own goal. Here is why, and what we changed.
It's the classic story of the shoemaker's children having no shoes. Tamer is designed to enable continuous delivery, orchestrated by AI agents, with a "2-day cadence" promise (a delivery or a post every two days). Yet, if you look at the dates on this blog, you will see a gaping 14-day hole.
Why? The irony is total: we were so busy building the perfect delivery tool that we forgot to deliver.
The Trap of Manual Supervision
In the beginning, every post on this blog was the result of a semi-automated workflow. An agent would draft, but a human had to validate every sentence, every commit, every Fly.io deployment.
As soon as the team had to focus on a major kernel release (v0.57.0), the blog became the variable that was adjusted. Human supervision became the bottleneck. We proved by absurdity that AI without autonomy is just a sophisticated secretarial tool, not a force multiplier.
What Changed: The Autonomous User Mode
This 14-day slip was the catalyst for a new feature: Autonomous Mode.
Instead of waiting for approval for every minor change, we learned to trust our "Critique Gates." We configured our agents so they could:
- Draft and self-critique according to a 7-point grid (see our next article).
- Deploy to a staging environment totally autonomously.
- Aggregate notifications so that a human only intervenes for the final "Publish."
It's not a matter of laziness; it's a matter of trust in the infrastructure.
Honesty Before Marketing
We could pretend this delay never happened. But agentic engineering is a field too new for "bullshit." If our cadence slipped, it's because our own processes were not yet mature enough for the autonomy we advocate.
Today, we are catching up (this post is part of a 5-article batch). But above all, we learned that cadence isn't a sprint; it's the result of a trusted infrastructure.
If you automate, don't supervise tasks. Supervise the rules.
Read also: